sensors and control interface methods based on ... · c. qiu et al.: sensors and control interface...

Received May 8, 2019, accepted July 1, 2019, date of publication July 8, 2019, date of current version July 26, 2019.

Digital Object Identifier 10.1109/ACCESS.2019.2927394

Sensors and Control Interface Methods Based onTriboelectric Nanogenerator in IoT ApplicationsCHUNKAI QIU 1, (Student Member, IEEE), FAN WU 1, (Student Member, IEEE),QIONGFENG SHI2, CHENGKUO LEE 2, AND MEHMET RASIT YUCE 1, (Senior Member, IEEE)1Department of Electrical and Computer Systems Engineering, Monash University, Melbourne, VIC 3800, Australia2Department of Electrical and Computer Engineering, National University of Singapore, Singapore 117576

Corresponding authors: Chengkuo Lee ([email protected]) and Mehmet Rasit Yuce ([email protected])

The work of M. R. Yuce was supported in part by the Australian Research Council Future Fellowships under Grant FT130100430, and inpart by the Travel Grant from the Faculty of Engineering, Monash University. This work was also supported by HIFES SeedFunding-2017-01 grant (R-263-501-012-133) ‘‘Hybrid Integration of Flexible Power Source and Pressure Sensors’’ at the NationalUniversity of Singapore.

ABSTRACT This paper presents a flexible sliding sensor that is capable of generating signals for wirelesscontrolling applications. The sensor is based on the coupling of triboelectrification and electrostatic inductioneffects. A signal processing circuit has been designed to process the alternating current (AC) signal generatedby the sensor, which is exerted to control a mobile robot in real time. Three triboelectric-based sensors havebeen designed in different sizes and shapes. Two of them are based on a peak-detection mechanism withdifferent control instructions formed from the varying number of peaks. The other octagonal-shaped sensoris based on an encoding method by changing the spacing between the strip electrodes. According to ourexperimental results, the triboelectric nanogenerator (TENG)-based sensors have high recognition accuracythat is higher than 97%. The flexibility of the sensor enables it to be attached to different devices, such asmobile phones, tablets, and laptops. The TENG-based sensors themselves can generate signals without anexternal power supply, which has a great advantage toward the Internet of Things (IoT) applications.

INDEX TERMS Triboelectric sensor, triboelectric nanogenerator (TENG), Internet of Things (IoT).

I. INTRODUCTIONThe Internet of Things (IoT) has attracted vast attention inrecent years, which enables the wireless connection and con-trolling of devices by the Internet [1], [2]. The IoT has beenused in applications for remote monitoring and controlling,such as healthcare [2], [3], smart home [4] and environmentalmonitoring [5], [6]. However, power consumption is a majorchallenge in developing IoT applications, because continuouspower is required by the sensor node to maintain the long-term connectivity [2]. Researchers have put a lot of effortinto the energy harvesting techniques for IoT applications,such as solar energy, kinetic energy and radio frequency (RF)energy [7]–[9]. Therefore, it is desirable to develop self-powered sensors for IoT applications that can reduce thepower consumption of sensor nodes.

Triboelectric nanogenerators (TENGs) convert mechan-ical energy into electric energy based on the triboelec-trification and electrostatic induction effects [10]–[12].

The associate editor coordinating the review of this manuscript andapproving it for publication was Vyasa Sai.

With the advantages of self-powered, cost-effective, sim-ple structure, easy fabrication, portable and high reliability[13], [14], TENGs are used in many different applications,such as energy harvesting [15]–[18], vibration sensing [19],impacts sensing [20], gas sensing [21], [22], humidity sens-ing [23], tracking system [24], and acceleration sensing [25].Because of its self-powered mechanism and low-cost,triboelectric-based sensors are gaining increasing popularityin IoT applications [26], [27].

With the rapid development of IoT technologies and sensortechnologies, there is a high demand for human-machineinterfaces (HMIs). Control interfaces are the essential compo-nents in the HMIs for handling human-machine interactions.Researchers have demonstrated great advantages of TENG-based HMIs, such as TENG-based touch sensors and key-boards. In [28], a transparent and flexible triboelectric sensingarray which is able to realise touch sensing, spatial mappingand trajectory is presented. The work in [29] presents awearable fabric keyboard for text-typing andmusical playing.A keyboard based on a novel resonant TENG is presentedin [30]. Each key can generate a unique range of oscillating

VOLUME 7, 2019 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ 92745

https://orcid.org/0000-0002-9729-8004

https://orcid.org/0000-0002-5011-5004

https://orcid.org/0000-0002-8886-3649

https://orcid.org/0000-0002-4802-391X

C. Qiu et al.: Sensors and Control Interface Methods Based on TENG in IoT Applications

frequencies. Therefore, a key press can be identified. Thesetriboelectric-based sensors have demonstrated great potentialin developing HMIs, including self-generated output, stableperformance, excellent durability, low-cost, and excellentflexibility. However, these triboelectric-array-based controlinterfaces have limitations in achieving higher resolution.Not only are additional sensor arrays required, but also addi-tional data acquisition units (DAQ) will be necessary. This isbecause each sensor unit in the array requires an independentdata acquisition channel.

In this paper, we propose a sensing circuit for detectingsignals generated from triboelectric-based sensors and cre-ating control signals for controlling remote robotic devicesvia a wireless XBee R© network for IoT applications. Thecontrol interface is based on a single-output-electrode. Thusonly one output data acquisition channel is required. Basedon the peak-detection and encoding algorithms proposed inthis work, higher control resolution can be achieved with-out increasing the number of output electrodes and DAQ,which demonstrates excellent potential in developing high-resolution control interfaces. Figure 1 shows the overviewof the triboelectric-based sensors system in the IoT appli-cations. Three sensors with a single output electrode arefabricated and tested in this work. Two of the sensors arebased on a peak-detection algorithm, where different controlinstructions will be generated according to the number ofpeaks detected. The third sensor is based on an encodingmethod, which is designed for the situation when numerousinstructions are required for controlling. To evaluate thefeasibility and performance of the proposed sensors and con-trolling circuit, a mobile robot with wireless live streamingvideo transmission system is designed and constructed fordemonstration.

FIGURE 1. System overview of the TENG-based IoT sensing system.

This paper is organized as follows: Section II describes themechanism of the triboelectric-based sensors and the detailsof the signal processing circuit design. Section III presents theexperimental results. The conclusion is given in section IV.

II. SYSTEM DESIGNThis section presents the implementation of the triboelectric-based sensor system towards the IoT applications. The systemconsists of 4 major parts: (1) triboelectric-based sensors; (2) asensor signal processing circuit; (3) a microcontroller (MCU)and a wireless transmitter and receiver for controlling the endnode; (4) a mobile robot with a camera and a live streamingvideo transmission system.

A. TRIBOELECTRIC-BASED SENSORSThe structure diagram and the design of triboelectric-basedsensors are shown in Figure 2. The triboelectric-based sensorsconsist of three layers. The top layer is a soft polytetrafluo-roethylene (PTFE) film with a thickness of 0.1 mm as thenegative triboelectric material [31], [32]. The middle layeris the electrode layer. It is a single-output-electrode. Theelectrode is made of aluminium (Al) foil strips (0.09 mmin thickness) and is deposited under the PTFE film as thesignal electrode. The bottom layer is a polydimethylsiloxane(PDMS) substrate. PDMS is used as the substrate becauseit is a flexible insulator. Due to its flexibility, the sensorcan be easily attached to different devices, such as mobilephones, tablets and laptops. Three different sensors have beenfabricated in this work as shown in Figure 2 (b-d). Each sensorhas a single-signal-output-electrode and is surrounded by Alfoil to the ground in order to minimise the influence of theunwanted electrostatic induction from the ambient environ-ment. The sensors depicted in Figure 2 (b) and (c) are basedon the peak-detection algorithm. As the finger slides from themiddle Al pad and across different number of electrodes indifferent directions, it will result in different number of outputpeaks corresponding to the number of electrodes. Therefore,direction sensing can be realised accordingly. Figure 2 (b)is an ID card-sized sensor with dimensions of 5.5 cm ×8.5 cm. The width of the Al electrode strip is 2 mm with3 mm spacing between two adjacent electrodes. The sensor,which is portable, is designed to be capable of attaching toID-cards or mobile phones. Figure 2 (c) is a tablet-sized sen-sor with dimensions of 13.5 cm× 23 cm. The width of the Alelectrode strip is 4mmwith 14mm spacing between two adja-cent electrodes. Compared with the ID-card size sensor, thistablet-sized sensor has a 14mm spacing between two adjacentelectrodes which allows the finger to slide on the sensordirectly. This sensor can be attached to tablets and laptopsfor the control purpose. Figure 2 (c) is an octagonal-shapedencoding-based sensor with the size of 27 cm × 30 cm.The width of the Al electrode strip is 4 mm. The encodingmethod is established with this shape by varying the spacingof strip electrodes.With the octagonal shape, eight codes havebeen embedded on the sensor for the sensing and controllingpurpose of eight directions.

Figure 3 illustrates the working mechanism of thetriboelectric-based self-powered sensors. The output signal isgenerated based on electrification and electrostatic inductioneffects [33]. Initially, there is no contact between the finger

92746 VOLUME 7, 2019


FIGURE 2. Top view of the three designs of triboelectric-based sensors with a single-output-electrode and the grounded outer shielding.(a) Structure design of the triboelectric-based sensors. (b) The ID card-sized sensor based on a peak-detection mechanism. (c) The tablet-sizedsensor based on a peak-detection mechanism. (d) The octagonal-shaped sensor based on an encoding mechanism.

FIGURE 3. Working mechanism of the triboelectric-based self-powered sensor. (a) The initial contact position of the sensor. (b)-(e) When fingerslides across the electrode, current is generated due to electrostatic induction effect. The states in (b)-(e) form a sliding operation cycle.

and sensor. Therefore, both the triboelectric sensor and thefinger are uncharged. After that, a hand wearing nitrile glovecontacts with the PTFE layer as depicted in Figure 3 (a).Because of the large difference in electron-attracting abilities,the PTFE film attracts the electrons from the nitrile glove,resulting in the net negatively charged PTFE surface and netpositively charged nitrile glove surface. Once the finger startsto slide on the PTFE surface, PTFE surface will attract moreelectrons from the nitrile glove. As a result, the triboelectriccharge density on the surface of PTFE layer and nitrile glovewill keep increasing and eventually becomes saturated anddistributes evenly on the surface. The negative triboelectriccharges on the PTFE surface will remain for a long period dueto the insulating property of the polymer material [17], [34].The negatively charged PTFE surface will induce positivecharges on the Al strip electrode as shown in Figure 3 (b)because of the electrostatic induction effect. When the fingerslides on the surface of PTFE film and approaches the elec-trode, positive charges on the glovewill attract electrons flow-ing from the ground to the Al electrode in order to balance out

the excessive positive charges on the electrode. The flow ofthe electrons from the ground to the Al electrode results ina current flowing from the electrode to the ground throughthe load. Thus a positive voltage will be generated on theload as depicted in Figure 3 (c). The positive voltage signalis induced until all the positive charges on the finger are bal-anced out by the induced electrons as shown in Figure 3 (d),where there is no current flow through the load. When thefinger slides away from the strip electrode, electrons willflow from the electrode to the ground, resulting in positivecharges re-induced on the electrode as shown in Figure 3(e). The flowing of electrons will result in a negative voltagegenerated on the load when referenced to the ground. Finally,as shown in Figure 3 (f), after the finger leaves the electrode,a new electrostatic equilibrium between the positive andnegative chargeswill be reached. The real-time output voltageand the quantity of charge transferred across the load whenthe finger is sliding across an Al electrode with a widthof 4 mm are depicted in Figure 4. When the positive voltagepeak is generated, electrons are flowing from the ground

VOLUME 7, 2019 92747


FIGURE 4. (a) The real-time output voltage and (b) charge transferredacross the load when the finger is sliding across an Al electrode with awidth of 4 mm.

to the electrode. Therefore, negative charges are transferredacross the load. While in the negative part of the outputvoltage, electrons flow back from the electrode to the ground,as a result, the charge transferred across the load returnsto zero. The result is consistent with the aforementionedworking mechanism in Figure 3. In this way, an alternatingcurrent (AC) signal is generated without any external powersupply. If this signal can be detected effectively, it can be usedfor controlling robotic devices and electronic system towardsthe IoT applications.

To investigate the performance and reliability of the pro-posed triboelectric-based sensor, for instance, the influenceof contact materials, contact areas, sliding speeds, and contactforces on the output of the sensor, the real-time output voltageand the transferred charge profile have been measured basedon a 4 mm wide Al electrode as shown in Figure 5.

Figure 5 (a) demonstrates the influence of different con-tact materials. Three circular-shaped contact materials withthe same size (diameter of 9 mm) are investigated, includ-ing Al foil, nylon textile, and a cotton textile coated withPoly (3,4-ethylenedioxythiophene): poly (styrene sulfonate)(PEDOT:PSS). PEDOT:PSS is amaterial that can increase thetriboelectric output level [26]. The sliding speed is controlledto be around 75 mm/s and the contact force is approximately2 N, which is almost the same speed and force as we slide onthe touchscreen of a mobile phone in daily life. The resultsindicate that with the larger relative position between the con-tact material and PTFE in triboelectric series, the higher out-put voltage is observed because the larger amount of chargeshave been induced and transferred. The PEDOT:PSS coatedtextile induces the largest amount of charges on the electrode,which leads to the highest output voltage in comparison withthe other two materials. Therefore, the PEDOT:PSS coatedtextile is utilised as the contact material in this work.

Figure 5 (b) investigates the effect of the contact areas onthe output by employing three circular-shaped PEDOT:PSScoated textile with diameters of 5 mm, 10 mm and 15 mm(the sliding speed is around 75 mm/s and the contact force isabout 2 N). As can be seen from the figure, the output voltageincreases as the contact area increases. This is because thelarger the contacting area is, the larger amount of transferredcharges (1Q) is induced.

The output voltage at three different levels of the slid-ing speed: slow (∼45 mm/s), normal (∼75 mm/s) and fast

FIGURE 5. Output voltage and the real-time charged transferred throughthe load under various testing conditions (x̄: Average of chargetransferred). (a) The influence of various contact materials: Aluminiumfoil, nylon textile and PEDOT:PSS coated textile. (b) The influence of thecontact areas of PEDOT:PSS coated textile with circular diameter of 5 mm,10 mm and 15 mm. (c) The influence of the sliding speeds: Slow∼45 mm/s, normal ∼75 mm/s and fast ∼200 mm/s. (d) The influence ofthe contact forces: ∼2 N, ∼5 N and ∼10 N.

(∼200 mm/s) have been investigated as shown in Figure 5 (c)(based on a PEDOT:PSS coated textile with diametersof 10 mm and the contact force is about 2 N). The amplitudeof the output voltage increases as the sliding speed increases.This is because the faster the sliding speed is, the faster theelectric charges are transferred. Therefore, it will result inhigher current through the load, given that almost the sameamount of charge is induced at different sliding speeds.

Figure 5 (d) demonstrates that the stronger the contactforce is, the larger the amount of triboelectric charge isinduced (based on a PEDOT:PSS coated textile with diam-eters of 10 mm and a sliding speed around 75 mm/s). Thisis because the stronger the contact force applied on the elec-trode, the closer the contact between the PEDOT:PSS coatedtextile and PTFE layer, leading to a larger effective contact

92748 VOLUME 7, 2019


FIGURE 6. Circuit diagram of the signal processing circuit for processing the voltage signal generated by the triboelectric-based sensors.

area and a better electrostatic induction. The results showthat the proposed sensor is robust under various contact forceworking conditions.

B. SIGNAL PROCESSING CIRCUITFigure 6 demonstrates the circuit diagram of the signal pro-cessing circuit. There are three major stages in the signalprocessing circuit, including the input stage, amplifier stageand comparator stage.

1) INPUT STAGEThe signal generated from the triboelectric-based sensor isan AC signal; however, a single power supply is used in thecircuit design. In the single supply system, the negative halfpart of the AC signal will be clipped. In order to obtain boththe positive and the negative parts of the AC signal, a DCoffset needs to be applied to the AC signal. The DC offsetis selected as one half of the power supply voltage to allowfor maximumAC signal scaling. The power supply voltage inthis system is 3.3 V. Thus the AC signal is shifted and centredat 1.65 V.

A voltage divider composed of two resistors is used to pro-vide the DC offset voltage. Because the triboelectric-basedsensors have high output impedance [13], the resistors withhigh resistance value are employed in the voltage divider inorder to provide a high input impedance of the signal pro-cessing circuit. Therefore, two 100 M� resistors are chosen.The biased AC signal is then followed by a voltage followerbecause it has a very high input impedance and will allow thestage 2 to receive sufficient voltage.

2) AMPLIFIER STAGEAnAC coupled inverting amplifier is implemented to amplifythe AC signal received from stage 1 [35]. The gain of theinverting amplifier is −2.13, which is set by the resistors

R3 and R4, assuming the AC coupling capacitor C2 is shortedin AC signal. The frequency response of the amplifier issimulated by the LTspice R© SPICE simulation software. Thelower cut-off frequency (fL) of the amplifier is 3.39 Hz.The simulation result is well matched with the measuredresult and the bandwidth is sufficient for the signal sensingapplication in this work.

3) COMPARATOR STAGEA non-inverting comparator is designed to compare theamplified signal with a threshold voltage. The comparator isused to identify a peak in the signal and filter out the noisein the signal. If the signal is higher than the threshold voltage(V+ > V−), the output will saturate toward the power supplyvoltage; otherwise, the output will saturate toward the ground.When the finger slides across the sensor, voltage peaks will begenerated. The peak voltage will be higher than the thresholdvoltage. Thus the output of the comparator will be 3.3 V.When the signal is below the threshold, the output will be 0 V.

The average signal output from stage 2 is the DC off-set 1.65 V, which is one half of the power supply voltage.Therefore, a voltage higher than 1.65 V needs to be chosen asthe threshold voltage. Two-thirds of the power supply voltage(2.2 V) is chosen in this work. The output of the comparatoris then sampled by the ADC embedded on the MCU at thefrequency of 500 Hz for data processing and controlling.

C. MICROCONTROLLER AND WIRELESSCOMMUNICATIONTheMCUused in the controlling circuit is ATmega328P fromAtmel R© [36]. It is a high performance and low power MCU.It has a built-in 8-channel 10-bit ADC which can be used tosample the output from the comparator. The MCU operatesat 3.3 V with a clock speed of 8 MHz.

VOLUME 7, 2019 92749


After the signal processing, the MCU will send corre-sponding instructions to control the mobile robot wirelesslythrough the XBee-PRO R© S3B module. The XBee-PRO R©

S3B module provides long-range wireless communicationbetween devices. It operates at 900 MHz and enables thelong-distance control of mobile robot. This is sufficient tocover the indoor and the outdoor controls.

D. THE DESIGN OF PCBA printed circuit board (PCB) of the signal processing circuithas been designed and fabricated as shown in Figure 7.The PCB is with the size of 4 cm wide and 8 cm long,which is suitable to be attached to mobile phones and IDcards. A rechargeable battery LIR2450 from Multicomp R© isemployed to power theMCU and the signal processing circuitto process the signal generated from the triboelectric-basedsensors.

FIGURE 7. The PCB design of the signal processing circuit to process thesignal generated from the triboelectric-based sensor.

E. AN APPLICATION EXAMPLE: MOBILE ROBOT CONTROLWITH LIVE STREAMING VIDEO SYSTEMFigure 8 depicts the block diagram of the robot control circuitfor processing the instructions sent by triboelectric sensors.As the XBee-PRO R© S3B receiver module receives signal,the MCU on the mobile robot will process the instructionsaccordingly, such as move forward, move backward, turnleft, and turn right. The MCU sends the control signal to amotor drive, which drives themotors of themobile robot. Two

FIGURE 8. Block diagram of robot control circuit for processing theinstructions sent by triboelectric sensors.

shaft encoders are equipped on the wheels to drive the robotstraight. Figure 9 depicts the design of the mobile robot.

FIGURE 9. Design of the mobile robot.

As shown in Figure 10, a camera is employed and theFirst Person View (FPV) live streaming video is transmittedby using the Turbowing 5.8 GHz transmitter to the EachineRC832 receiver. The live streaming video is shown on anLCD screen. The transmitter and receiver operate at 5.8 GHz,and their transmission range is up to 100 m.

FIGURE 10. Live streaming video transmission system on the robot.

III. EXPERIMENTAL RESULTSTo evaluate the feasibility of the triboelectric-based sensorsand the performance of the signal processing circuit, threesensors are fabricated and tested. The first two are based onthe peak-detection method and the third one is based on theencoding method. The first one is an ID card-sized sensor,which can be attached to the mobile phones for controlling“things” wirelessly; the second one is a tablet-sized sensor,which can be attached to the tablets or the walls; the third oneis an octagonal-shaped sensor.

Figure 11 shows the procedure of the peak-detectionmethod with noise peak rejection. The figure shows a 2-peaksignal with a noise peak. The signal is sampled by the ADCat a frequency of 500 Hz in order to capture the rising and thefalling edges of the signal effectively. In this example, 6 edges(E1 to E6) are captured and the corresponding time (Ti) isrecorded as well. Thus the duration of each peak (Di) can becalculated. The noise peak (D2) will be removed because itsduration is less than the predefined threshold duration (Tth).In order to detect the number of peaks accurately, it is nec-essary to choose an appropriate Tth value for noise peakrejection. Based on experimental results, the duration of each

92750 VOLUME 7, 2019


FIGURE 11. The peak-detection method utilized in the signal processingcircuit.

noise peak generated in ID card-sized sensor is less than 5ms,and less than 10 ms in both tablet-sized and octagonal-shapedsensors. The duration for an effective peak generated bysliding across an electrode is much longer than a noise peak.Therefore, the noise peak is rejected by setting an appropriatethreshold value (Tth) in time domain. Based on this method,the 2-peak signal can be detected successfully. The softwarealgorithm diagram of the peak-detection method is illustratedin Figure 12.

FIGURE 12. Software algorithm flowchart of the peak-detection methodfor triboelectric-based sensors.

A. RESULTS OF ID CARD-SIZED SENSORTo avoid sweaty hands affecting the performance of the sen-sor, and to overcome the limitation that the fingertip is largerthan the spacing between the strip electrodes, a glove withfinger cot is deployed as shown in Figure 13 (a) and (b).A textile coated with PEDOT:PSS is attached to the tip offinger cot to increase the triboelectric output. A sliding startsfrom the middle grounded Al pad and ends at the outergrounded shield is illustrated in Figure 13 (c) and (d).

Figure 14 shows the output voltage signals of the ID card-sized sensor at three different stages of the signal processingcircuit under normal (∼75 mm/s) sliding speed. The output

FIGURE 13. The ID card-sized sensor system. (a) and (b) are the side viewof finger cot. (c) a sliding starts from the middle Al pad; (d) a sliding endsat the outer Al shield.

FIGURE 14. The generated output signal of the ID card-sizedtriboelectric-based sensor at different stages of the signal processingcircuit when different instructions are formed under normal sliding speed(∼75 mm/s): (a) instruction 1 (1 peak); (b) instruction 2 (2 peaks);(c) instruction 3 (3 peaks); (d) instruction 4 (4 peaks).

signal of stage 1, which is the AC signal generated by thetriboelectric-based sensor, shifted up by 1.65 V to the centreof the supply voltage. The output of the inverting amplifieris centred at 1.65 V as well, while the AC voltage has beenamplified by the gain of 2.13. As can be seen from the plot,some peaks are clipped. However, clipping is not a problemin this work, because the information of the number of peaksis not lost. The peaks are transformed to square pulse by thecomparator in stage 3, and the number of peaks can be easilyidentified at this stage. When different numbers of peaksare detected, corresponding instructions will be sent by theXBee R© transmitter module to control the action of the mobile

VOLUME 7, 2019 92751


FIGURE 15. Instructions are formed to control the movement of robotsbased on the signal generated from ID card-sized triboelectric-basedsensor.

FIGURE 16. The generated output signal at (a) slow (∼45 mm/s) and(b) fast (∼200 mm/s) sliding speeds.

robot according to the instructions given in Figure 15. Forexample, when the participant slides upwards, one peak willbe generated and detected. The XBee R© transmitter modulewill then transmit ‘‘Forwards’’ instruction to the robot. Therobot receives the instruction and then moves forward. Thesystem is capable of detecting peaks under various slidingspeeds as shown in Figure 16. With higher sliding speed,the amplitude of the output voltage will increase, and thewidth of the output voltage pulse will be narrower due to theshorter time of pulse generation. To evaluate the accuracy ofthe system, experiments have been conducted on 7 subjectsto validate the accuracy of the system. In the experiment,the subject wears the glove and triggers each instruction100 times. The recognition accuracy equals the number ofcorrect recognition dividing the total number of instruc-tion triggering. Table 1 is the recognition accuracy of the

TABLE 1. Recognition accuracy of the ID card-sized sensor (outof 100 times measurements).

ID card-sized sensor. The result suggests that a high recog-nition accuracy has been achieved, which is above 99% onaverage.

B. RESULTS OF TABLET-SIZED SENSORFigure 17 (a) and (b) illustrate a sliding posture which startsfrom the middle grounded Al pad and ends at the outergrounded shield. When the finger is sliding across four elec-trode strips, a 4-peak pulse will be generated. The spacingbetween the strip electrodes is 14 mm, which enables thefinger to slide on the sensor directly. A glove with a textilecoated with PEDOT:PSS is employed to increase the tribo-electric output [26] as shown in Figure 17 (c). Figure 18shows the output signals of the tablet-sized sensor at threedifferent stages of the signal processing circuit when dif-ferent instructions are formed. The system is able to detectpeaks effectively under various sliding speeds. The numberof correct recognition dividing the total number of instructiontriggering is the recognition accuracy. Table 2 shows therecognition accuracy is over 99% on average from the resultsof 7 subjects.

FIGURE 17. The tablet-sized sensor system. (a) a sliding starts from themiddle Al pad; (b) a sliding ends at the outer Al shield; (c) a glove with atextile coated with PEDOT:PSS.

FIGURE 18. The generated output signal of the tablet-sizedtriboelectric-based sensor at different stages of the signal processingcircuit when different instructions are formed under normal sliding speed(∼75 mm/s): (a) instruction 1 (1 peak); (b) instruction 2 (2 peaks);(c) instruction 3 (3 peaks); (d) instruction 4 (4 peaks).

92752 VOLUME 7, 2019


TABLE 2. Recognition accuracy of tablet-sized sensor (out of 100 timesmeasurements).

C. RESULTS OF OCTAGONAL-SHAPED SENSORThe octagonal-shaped sensor is encoded with 8 decimal num-bers represented in binary format. Each code represents arotate instruction of the mobile robot; hence a 360◦ landscapecan be captured by the camera on the robot. The encodingmethod is represented with the octagonal-shaped sensor asdepicted in Figure 19. A positive peak in the output will berecognised as digit number “1”, and zero volts in the outputwill be recognised as digit number “0”.When the finger slidesacross an electrode strip, a positive peak will be generated inthe output. Therefore, an electrode strip is used to representthe digit number “1”. All of the codes begin with a startbit “1” and end with a stop bit “1”.

FIGURE 19. Encoding method of the octagonal-shaped sensor.

Figure 20 illustrates the decoding method of the system.After the MCU receives the input signal, the software pro-gram will calculate the number of peaks between the start bitand the stop bit. For instance, 1 peak is detected between thestart bit and the stop bit in Figure 20 (a). Then the systemwill calculate the total off-peak pulse duration (Toff−peak ),which is the summation of each individual off-peak dura-tion (Ti), and the number of bits represented by the off-peakduration (Bitsoff−peak ). Each code has a total length of 7 bits.In Figure 20 (a), 1 peak is detected between the start bitand the stop bit. Therefore, there are 6 digits of “0” dur-ing Toff−peak . The average duration for a single digit“0” (Taverage) is calculated by the equation 1, and the ratioof each off-peak duration Ti_ratio is calculated by equation 2,where Ti is an individual off-peak duration. The Ti_ratio isthen compared with a margin given in the Table 3 to con-vert the ratio to integer module values. For example, T1in Figure 20 (a) will be converted to 5 digits “0” and T2 will be

FIGURE 20. Decoding method of the octagonal-shaped sensor.

TABLE 3. Margin table to convert pulse width to the number of digits.

converted to 1 digit “0”. In this way, the code “0000010” canbe decoded and a direction instruction to turn the robot to theeast (E) direction will be sent. The margin of the ratio is givenby the 95% confidence interval based on a preliminary testingdata set of 200 measurements. The margin will allow non-uniform sliding speed across the strip electrodes and the codestill can be recognised successfully. However, themargin onlyallows a certain level of non-uniform sliding speed. If thespeed exceeds the margin level, an error will occur. Figure 21shows the output signals of the sensor when sliding acrossdifferent codes. Table 4 shows the recognition accuracy isabove 97% on average from the results of 3 subjects.

Taverage =Toff−peakBitsoff−peak

(1)

Ti_ratio =Ti

Taverage(2)

VOLUME 7, 2019 92753


FIGURE 21. The generated output signal of the octagon-shapedtriboelectric-based sensor at different stages of the signal processingcircuit when different instructions are formed. (a)-(h) represent8 different instructions.

TABLE 4. Recognition accuracy of the octagonal-shaped sensor (outof 100 times measurements).

Figure 22 illustrates an example of robot movements whendifferent instructions are triggered. Figure 22 (a) shows thestarting position of the robot. Then the code ‘010’ is triggered,

the robot turns right by 45◦ to the north-east (NE) directionas shown in Figure 22 (b). After that, if the code ‘000’ istriggered, the robot will make a 90◦ turn to the south-east (SE)direction as shown in Figure 22 (c). Lastly, the robot will turnto the north-west (NW) when the code ‘100’ is triggered.With the control of the octagonal-shaped sensor, the robotcan rotate in 8 directions. In this way, a 360◦ panorama viewcan be obtained by the live streaming video system on therobot. A video demonstrates the real-time control of the robotrotation can be found in the Supporting Material Video 1.

FIGURE 22. An example of robot movements. (a) starting position;(b) when code ‘010’ is triggered; (c) then code ‘000’ is triggered;(d) lastly, code ‘100’ is triggered.

D. APPLICATION EXAMPLESA prototype device for controlling the mobile robot isshown in Figure 23. Figure 23 (a) depicts a smaller IDcard-sized sensor is attached to the top layer of the PCB

FIGURE 23. A prototype of the proposed device. (a) A smaller version ofthe ID card-sized sensor; (b) a highly stretchable and flexible siliconerubber substrate; (c) the prototype sensor is attached to a mobile phone;(d) the prototype sensor is attached to an ID card.

92754 VOLUME 7, 2019


TABLE 5. Comparison of different human-machine control interfaces based on TENG.

as the controlling sensor. The electronic components layerof the PCB is embedded on a silicone rubber substrateas shown in Figure 23 (b). The substrate is made fromEcoflexTM 00-10. It has a length of 8 cm, a width of 4 cm,and a height of 1 cm. EcoflexTM rubber is very soft, stretch-able and flexible. The good adhesion of the EcoflexTM

rubber enables the PCB to be attached to many devices,such as the mobile phones and the ID-cards as shownin Figure 23 (c) and (d). A video shows the real-time controlof the robot movements by this prototype can be found in theSupporting Material Video 2.

Table 5 summaries some recent TENG-based control inter-faces such as touch sensors and keyboards. Compared withother works, the sensors proposed in this work have single-output-electrode, and higher control resolution without addi-tional data acquisition channel. The signal is processed on theMCU, and control instructions can be sent wirelessly via theportable device.

IV. CONCLUSIONThis paper presents a flexible, single-output-electrodetriboelectric-based sensor that can be used for controllingremote IoT devices. Three sensors and control interfaces havebeen presented. Two of them are based on a peak-detectionmechanism in which a different number of peaks in the outputsignal is used to form different control instructions. The otheroctagonal-shaped sensor is based on an encoding method.With this octagon shape, eight codes can be encoded onthe sensor and form eight control instructions. The signalprocessing circuit and the sensing control mechanism withdecoding algorithms are designed for the proposed triboelec-tric sensors, which are used for real-time controlling of amobile robot with wireless live streaming video. The pro-posed sensors can generate multiple voltage peak signals witha single-output-electrode, which demonstrates a great poten-tial to increase the controlling instructions without increasingthe number of output electrodes. High recognition accuracy

of above 97% on average are achieved from several subjects’experimental tests. This work demonstrates the possibilitythat the TENG-based sensors can be used for real-time con-trolling of embedded systems. The flexibility, self-powered,single-electrode-output, and low-cost of the TENG-basedsensor show great advantages toward remote controlling ofIoT devices and systems.

REFERENCES[1] W. Yu, F. Liang, X. He, W. G. Hatcher, C. Lu, J. Lin, and X. Yang,

‘‘A survey on the edge computing for the Internet of Things,’’ IEEE Access,vol. 6, pp. 6900–6919, 2018.

[2] T. Wu, F. Wu, and J.-M. Redouté, and M. R. Yuce, ‘‘An autonomous wire-less body area network implementation towards IoT connected healthcareapplications,’’ IEEE Access, vol. 5, pp. 11413–11422, 2017.

[3] S. B. Baker, W. Xiang, and I. Atkinson, ‘‘Internet of Things for smarthealthcare: Technologies, challenges, and opportunities,’’ IEEE Access,vol. 5, pp. 26521–26544, 2017.

[4] B. L. R. Stojkoska and K. V. Trivodaliev, ‘‘A review of Internet of Thingsfor smart home: Challenges and solutions,’’ J. Cleaner Prod., vol. 140,no. 3, pp. 1454–1464, 2017.

[5] S. Fang, L. D. Xu, Y. Zhu, J. Ahati, H. Pei, J. Yan, and Z. Liu, ‘‘An inte-grated system for regional environmental monitoring and managementbased on Internet of Things,’’ IEEE Trans. Ind. Informat., vol. 10, no. 2,pp. 1596–1605, May 2014.

[6] F. Wu, C. Rudiger, and M. R. Yuce, ‘‘Design and field test of anautonomous IoT WSN platform for environmental monitoring,’’ in Proc.27th Int. Telecommun. Netw. Appl. Conf. (ITNAC), Nov. 2017, pp. 1–6.

[7] T. Wu and J.-M. Redouté, and M. R. Yuce, ‘‘A wireless implantable sensordesign with subcutaneous energy harvesting for long-term IoT healthcareapplications,’’ IEEE Access, vol. 6, pp. 35801–35808, 2018.

[8] P. Kamalinejad, C. Mahapatra, Z. Sheng, S. Mirabbasi, V. C. M. Leung,and Y. L. Guan, ‘‘Wireless energy harvesting for the Internet of Things,’’IEEE Commun. Mag., vol. 53, no. 6, pp. 102–108, Jun. 2015.

[9] P. Lynggaard, ‘‘A self-supporting wireless IoT node that uses kineticenergy harvesting,’’ in Proc. Internet of Things Bus. Models, Users, Netw.,Nov. 2017, pp. 1–6.

[10] Z. L. Wang, ‘‘Triboelectric nanogenerators as new energy technology forself-powered systems and as active mechanical and chemical sensors,’’ACS Nano, vol. 7, no. 11, pp. 9533–9557, 2013.

[11] Q. Shi, T. He, and C. Lee, ‘‘More than energy harvesting—Combining tri-boelectric nanogenerator and flexible electronics technology for enablingnovel micro-/nano-systems,’’ Nano Energy, vol. 57, pp. 851–871,Mar. 2019.

VOLUME 7, 2019 92755


[12] Y. Su, J. Chen, Z. Wu, and Y. Jiang, ‘‘Low temperature dependence oftriboelectric effect for energy harvesting and self-powered active sensing,’’Appl. Phys. Lett., vol. 106, no. 1, 2015, Art. no. 013114.

[13] Y. Xie, S. Wang, L. Lin, Q. Jing, Z.-H. Lin, S. Niu, Z. Wu, and Z. L. Wang,‘‘Rotary triboelectric nanogenerator based on a hybridized mechanism forharvesting wind energy,’’ ACS Nano, vol. 7, no. 8, pp. 7119–7125, 2013.

[14] L. Chen, Q. Shi, Y. Sun, T. Nguyen, C. Lee, and S. Soh, ‘‘Controllingsurface charge generated by contact electrification: Strategies and appli-cations,’’ Adv. Mater., vol. 30, no. 47, 2018, Art. no. 1802405.

[15] Y. Feng, L. Zhang, Y. Zheng, D. Wang, F. Zhou, and W. Liu, ‘‘Leavesbased triboelectric nanogenerator (TENG) and TENG tree for wind energyharvesting,’’ Nano Energy, vol. 55, pp. 260–268, Jan. 2019.

[16] K. Xia, Z. Zhu, H. Zhang, C. Du, J. Fu, and Z. Xu, ‘‘Milk-based tribo-electric nanogenerator on paper for harvesting energy from human bodymotion,’’ Nano Energy, vol. 56, pp. 400–410, Feb. 2019.

[17] Y. Su, Y. Yang, X. Zhong, H. Zhang, Z. Wu, Y. Jiang, and Z. L. Wang,‘‘Fully enclosed cylindrical single-electrode-based triboelectric nanogen-erator,’’ ACS Appl. Mater. Interfaces, vol. 6, no. 1, pp. 553–559, 2014.

[18] Y. Su, X. Wen, G. Zhu, J. Yang, J. Chen, P. Bai, Z. Wu, Y. Jiang, andZ. L.Wang, ‘‘Hybrid triboelectric nanogenerator for harvesting water waveenergy and as a self-powered distress signal emitter,’’ Nano Energy, vol. 9,pp. 186–195, Oct. 2014.

[19] Q. Liang, Z. Zhanga, X. Yan, Y. Gu, Y. Zhao, G. Zhang, S. Lu, Q. Liao, andY. Zhang, ‘‘Functional triboelectric generator as self-powered vibrationsensor with contact mode and non-contact mode,’’ Nano Energy, vol. 14,pp. 209–216, May 2015.

[20] C. Garciaa, I. Trendafilova, and J. S. D. Rio, ‘‘Detection and measurementof impacts in composite structures using a self-powered triboelectric sen-sor,’’ Nano Energy, vol. 56, pp. 443–453, Feb. 2019.

[21] Y. Su, G. Xie, H. Tai, S. Li, B. Yang, S. Wang, Q. Zhang,H. Du, H. Zhang, X. Du, and Y. Jiang, ‘‘Self-powered room temperatureNO2 detection driven by triboelectric nanogenerator under UV illumina-tion,’’ Nano Energy, vol. 47, pp. 316–324, May 2018.

[22] K. Zhao, G. Gu, Y. Zhang, B. Zhang, F. Yang, L. Zhao, M. Zheng,G. Cheng, and Z. Du, ‘‘The self-powered CO2 gas sensor based on gasdischarge induced by triboelectric nanogenerator,’’ Nano Energy, vol. 53,pp. 898–905, Nov. 2018.

[23] Y. Su, G. Xie, S. Wang, H. Tai, Q. Zhang, H. Du, H. Zhang, X. Du, andY. Jiang, ‘‘Novel high-performance self-powered humidity detectionenabled by triboelectric effect,’’ Sens. Actuators B, Chem., vol. 251,pp. 144–152, Nov. 2017.

[24] Y. Su, G. Zhu, W. Yang, J. Yang, J. Chen, Q. Jing, Z. Wu, Y. Jiang, andZ. L. Wang, ‘‘Triboelectric sensor for self-powered tracking of objectmotion inside tubing,’’ ACS Nano, vol. 8, no. 4, pp. 3843–3850, 2014.

[25] Q. Shi, H. Wu, H. Wang, H. Wu, and C. Lee, ‘‘Self-powered gyroscopeball using a triboelectric mechanism,’’ Adv. Energy Mater., vol. 7, no. 22,2017, Art. no. 1701300.

[26] T. He, Q. Shi, H. Wang, F. Wen, T. Chen, J. Ouyang, and C. Lee, ‘‘Beyondenergy harvesting-multi-functional triboelectric nanosensors on a textile,’’Nano Energy, vol. 57, pp. 338–352, Mar. 2019.

[27] T. Chen, Q. Shi, K. Li, Z. Yang, H. Liu, L. Sun, J. A. Dziuban, andC. Lee, ‘‘Investigation of position sensing and energy harvesting of aflexible triboelectric touch pad,’’Nanomaterials, vol. 8, no. 8, p. 613, 2018.

[28] Z. Yuan, T. Zhou, Y. Yin, R. Cao, C. Li, and Z. L. Wang, ‘‘Transparent andflexible triboelectric sensing array for touch security applications,’’ ACSNano, vol. 11, no. 8, pp. 8364–8369, 2017.

[29] S.-B. Jeon, S.-J. Park, W.-G. Kim, I.-W. Tcho, I.-K. Jin, J.-K. Han, D. Kim,and Y.-K. Choi, ‘‘Self-powered wearable keyboard with fabric based tribo-electric nanogenerator,’’ Nano Energy, vol. 53, pp. 596–603, Nov. 2018.

[30] W. Yin, Y. Xie, J. Long, P. Zhao, J. Chen, J. Luo, X. Wang, and S. Dong,‘‘A self-power-transmission and non-contact-reception keyboard based ona novel resonant triboelectric nanogenerator (R-TENG),’’ Nano Energy,vol. 50, pp. 16–24, Aug. 2018.

[31] Z. L. Wang, J. Chen, and L. Lin, ‘‘Progress in triboelectric nanogeneratorsas a new energy technology and self-powered sensors,’’ Energy Environ.Sci., vol. 8, pp. 2250–2282, May 2015.

[32] A. Diaz and R.M. Felix-Navarro, ‘‘A semi-quantitative tribo-electric seriesfor polymeric materials: The influence of chemical structure and proper-ties,’’ J. Electrostatics, vol. 62, no. 4, pp. 277–290, 2004.

[33] S. Niu, S. Wang, Y. Liu, Y. S. Zhou, L. Lin, Y. Hu, K. C. Pradel, andZ. L. Wang, ‘‘A theoretical study of grating structured triboelectric nano-generators,’’ Energy Environ. Sci., vol. 7, no. 7, pp. 2339–2349, 2014.

[34] F. Saurenbach, D. Wollmann, B. D. Terris, and A. F. Diaz, ‘‘Forcemicroscopy of ion-containing polymer surfaces: Morphology and chargestructure,’’ Langmuir, vol. 8, no. 4, pp. 1199–1203, 1992.

[35] AC Coupled, Single-Supply, Inverting and Non-Inverting Amplifier Refer-ence Design, Texas Instruments, Dallas, TX, USA, 2015.

[36] Atmel Atmega328p Datasheet, ATMEL Corp., San Jose, CA, USA, 2011.[37] X. Wang, H. Zhang, L. Dong, X. Han, W. Du, and J. Zhai, ‘‘Self-powered

high-resolution and pressure-sensitive triboelectric sensor matrix for real-time tactile mapping,’’ Adv. Mater., vol. 28, no. 15, pp. 2896–2903, 2016.

[38] S. Wang, L. Lin, and Z. L. Wang, ‘‘Triboelectric nanogenerators as self-powered active sensors,’’ Nano Energy, vol. 11, pp. 436–462, Jan. 2015.

CHUNKAI QIU received the B.E. degree fromMonash University, Australia, in 2017, where heis currently pursuing the Ph.D. degree with theDepartment of Electrical and Computer SystemsEngineering. His current research interests includewearable devices and triboelectric nanogenerators.

FAN WU received the B.E. degree from MonashUniversity, in 2015, where he is currently pursu-ing the Ph.D. degree in electrical and computersystems engineering. He was a Research Assistantwith the Engineering Department, from 2015 to2017. His current research interests include wire-less sensor networks, wearable sensors, energyharvesting, triboelectric nanogenerators, and theIoT innovations.

QIONGFENG SHI received the B.Eng. degreefrom the Department of Electronic Engineeringand Information Science, University of Scienceand Technology of China (USTC), in 2012, andthe Ph.D. degree from the Department of Electricaland Computer Engineering, National Universityof Singapore (NUS), in 2018. He is currently aResearch Fellowwith the Department of Electricaland Computer Engineering, National University ofSingapore. His research interests include energy

harvesters, triboelectric nanogenerators, self-powered sensors, andwearable/implantable electronics.

CHENGKUO LEE received the Ph.D. degree inprecision engineering from The University ofTokyo, in 1996. In 2001, he co-founded AsiaPacific Microsystems, Inc., where he was theVice President. From 2006 to 2009, he was aSenior Member of the Technical Staff with theInstitute ofMicroelectronics, A-STAR, Singapore.He is currently the Director of the Center forIntelligent Sensors and MEMS, and an AssociateProfessor with the Department of Electrical and

Computer Engineering, National University of Singapore, Singapore. Hehas contributed to more than 300 international conference papers, extendedabstracts, and 280 peer-reviewed international journal articles.

92756 VOLUME 7, 2019


MEHMET RASIT YUCE received the M.S. degreein electrical and computer engineering from theUniversity of Florida, Gainesville, FL, USA,in 2001, and the Ph.D. degree in electrical andcomputer engineering from North Carolina StateUniversity (NCSU), Raleigh, NC, USA, in 2004.He was a Postdoctoral Researcher with the Elec-trical Engineering Department, University ofCalifornia at Santa Cruz, in 2005. He was anAcademic Member with the School of Electri-

cal Engineering and Computer Science, University of Newcastle, NSW,Australia, until 2011. In 2011, he joinedMonash University, Australia, wherehe is currently an Associate Professor with the Department of Electricaland Computer Systems Engineering. His research interests include wearabledevices, the Internet of Things (IoT) for healthcare, wireless implantable

telemetry, wireless body area networks (WBAN), bio-sensors, integratedcircuit technology dealing with digital, analog, and radio frequency circuitdesigns for wireless, biomedical, andRF applications. He has publishedmorethan 150 technical articles in the above areas.

He received the NASAGroupAchievement Award, in 2007 for developingan SOI transceiver. He received the Best Journal Paper Award, in 2014 fromthe IEEE Microwave Theory and Techniques Society (MTTS). He receivedthe Research Excellence Award from the Faculty of Engineering and BuiltEnvironment, University of Newcastle, in 2010. He has authored the booksWireless Body Area Networks (2011) and Ultra-Wideband and 60 GHzCommunications for Biomedical Applications (2013). He is a Topical Editorof the IEEE SENSORS JOURNAL, an Associate Editor-in-Chief of Sensors, anda Guest Editor of the IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS,in 2015.

VOLUME 7, 2019 92757

sensors and control interface methods based on ... · c. qiu et al.: sensors and control interface...

Documents